GPU Comparison

NVIDIA GeForce GT 640 and Intel UHD Graphics 610 occupy opposite ends of the GPU spectrum: the former is a dedicated discrete card from 2012 built on Kepler, while the latter is a 2020 integrated processor graphics solution from Intel’s Comet Lake generation. Benchmark data shows the GT 640 ahead in every head-to-head test, but the UHD 610 counters with a much newer feature set and dramatically lower power draw. This analysis quantifies those differences using only the provided figures.

Head-to-Head Benchmarks

The GT 640 wins both available benchmark comparisons, and the margins are substantial. In Geekbench OpenCL, the NVIDIA card scores 3,736 against the Intel iGPU’s 2,692, a 38.8% advantage. The Vulkan test shows a similar gap: 3,809 for the GT 640 versus 2,811 for the UHD 610, a 35.5% lead. These are not marginal wins; they represent a full performance tier between the two products.

The GT 640’s average benchmark score of 3,210 places it at the 20th percentile among all GPUs, while the UHD 610 averages 2,752 and sits at the 18th percentile. Although the percentile gap appears small, the raw score difference is 458 points, roughly 16.6% of the Intel part’s average. The OpenCL delta is even larger than the average, indicating that the GT 640’s lead grows in compute-oriented workloads.

Looking at the nearest rivals for context, the GT 640’s average score sits 0.9% above the Intel Arc Pro B60 (3,182) and 1.5% above the NVIDIA Quadro P1000 (3,163). It trails the GeForce 920M by 2.3% (3,287) and the GT 730M by 3.2% (3,316). The UHD 610, meanwhile, is 1.1% below the GeForce 820M (2,784) and 1.5% below the RTX 3060 Ti GDDR6X (2,795), but 2.3% above the RTX 5070 SUPER (2,690). These rival deltas highlight that both products cluster near the bottom of the performance distribution, yet the GT 640 consistently lands on the higher side of that cluster.

The Vulkan result is particularly telling. The GT 640 scores 3,809, which is 28.9% higher than its own OpenCL score of 3,736. The UHD 610 shows a smaller API gap: 2,811 in Vulkan versus 2,692 in OpenCL, a 4.4% improvement. This suggests the NVIDIA architecture scales better with modern API overhead, even though the GT 640 predates the UHD 610 by eight years.

No benchmark favors the Intel part. The wins tally is 2–0, and the deltas (38.8% and 35.5%) are consistent across both APIs. For any workload represented by these tests, the GT 640 is the faster GPU.

FAQ

Q: Which GPU has the higher average benchmark score?

A: The NVIDIA GeForce GT 640 averages 3,210 points, which is 458 points higher than the Intel UHD Graphics 610’s average of 2,752. The GT 640 also holds a higher percentile rank at 20 versus 18.

Q: How much faster is the GT 640 in OpenCL?

A: The GT 640 scores 3,736 in Geekbench OpenCL versus 2,692 for the UHD 610, a 38.8% advantage. This is the largest delta between the two parts in any test.

Q: Does the Intel UHD 610 win any benchmark?

A: No. The UHD 610 loses both head-to-head tests, with the GT 640 winning OpenCL by 38.8% and Vulkan by 35.5%. The wins tally is 2 for NVIDIA and 0 for Intel.

Q: How does the UHD 610 compare to its nearest rivals?

A: The UHD 610’s average score of 2,752 is 1.1% below the NVIDIA GeForce 820M (2,784) and 1.5% below the RTX 3060 Ti GDDR6X (2,795). It is 1.7% below the AMD Radeon 550 (2,801) but 2.3% above the RTX 5070 SUPER (2,690).

Q: What is the GT 640’s position among its closest competitors?

A: The GT 640’s average of 3,210 is 0.9% above the Intel Arc Pro B60 (3,182) and 1.5% above the Quadro P1000 (3,163). It trails the GeForce 920M (3,287) by 2.3% and the GT 730M (3,316) by 3.2%.

Q: Which GPU supports a higher DirectX version?

A: The Intel UHD Graphics 610 supports DirectX 12 (12_1), while the NVIDIA GT 640 supports DirectX 12 (11_0). The Intel part also supports Vulkan 1.3 versus the GT 640’s Vulkan 1.2.175.

Architecture Differences

The two GPUs come from fundamentally different design philosophies and manufacturing eras. The GT 640 uses NVIDIA’s Kepler architecture on a 28 nm TSMC process, while the UHD 610 employs Intel’s Generation 9.5 architecture on a 14 nm+++ process. The GT 640 is a discrete GPU built around the GK107 chip, whereas the UHD 610 is an integrated processor graphics solution using the Comet Lake GT1 die.

The compute resources diverge sharply. The GT 640 packs 384 shading units, 32 texture mapping units (TMUs), and 16 raster output units (ROPs). The UHD 610 has 96 shading units, 12 TMUs, and only 2 ROPs. That is a 4x advantage in shading units and TMUs for NVIDIA, and an 8x advantage in ROPs. These ratios explain the GT 640’s dominance in fill-rate-limited workloads.

Memory architecture is another major divider. The GT 640 has 2 GB of dedicated DDR3 memory on a 128-bit bus, delivering 28.51 GB/s of bandwidth. The UHD 610 uses system shared memory with a system-dependent bandwidth figure. The GT 640’s memory clock runs at 891 MHz with 1782 Mbps effective speed; the UHD 610 has no dedicated memory clock.

The die itself tells a story of different priorities. The GT 640’s GK107 contains 1,270 million transistors on a 118 mm² die, yielding a density of 10.8 million transistors per mm². Intel does not disclose transistor count or die size for the Comet Lake GT1, but the 14 nm+++ process and integrated nature suggest a much smaller dedicated GPU area. The GT 640’s transistor density is a direct consequence of the mature 28 nm planar process, while Intel’s 14 nm+++ is a refined FinFET technology.

Feature support flips the generational script. The UHD 610 supports DirectX 12 (12_1), Vulkan 1.3, and OpenGL 4.6. The GT 640 supports DirectX 12 (11_0), Vulkan 1.2.175, and OpenGL 4.6. While both handle OpenGL 4.6, the Intel part has a higher feature level in DirectX and a newer Vulkan revision. The GT 640 also has dedicated display outputs (1x DVI, 1x HDMI 1.4a, 1x DisplayPort 1.2), while the UHD 610’s outputs are motherboard dependent.

Specification Differences

The two GPUs differ in nearly every measurable specification. Process node: 28 nm (TSMC) for the GT 640 versus 14 nm+++ (Intel) for the UHD 610. Transistor count: 1,270 million versus not disclosed. Die size: 118 mm² versus not disclosed. Transistor density: 10.8M / mm² versus not disclosed.

Clock behavior is fundamentally different. The GT 640 lists only a memory clock of 891 MHz (1782 Mbps effective), with no base or boost clock provided. The UHD 610 has a base clock of 350 MHz and a boost clock of 1050 MHz, with no separate memory clock since it uses system memory.

Memory specifications are split between dedicated and shared. The GT 640 has 2 GB of DDR3 on a 128-bit bus with 28.51 GB/s bandwidth. The UHD 610 has system shared memory with system shared type, bus width, and system dependent bandwidth.

Compute rates favor the GT 640 overwhelmingly. Pixel rate: 7.216 GPixel/s versus 2.100 GPixel/s. Texture rate: 28.86 GTexel/s versus 12.60 GTexel/s. FP32: 692.7 GFLOPS versus 201.6 GFLOPS. The GT 640 has no FP16 figure, while the UHD 610 delivers 403.2 GFLOPS FP16 (2:1 ratio).

Power and physical specs are starkly different. The GT 640 has a 65 W TDP, single-slot width, no power connectors, and a suggested PSU of 250 W. The UHD 610 has a 15 W TDP and is an IGP with no slot width, no power connectors, and no suggested PSU. The GT 640 measures 145 mm (5.7 inches) in length; the UHD 610 has no physical dimensions since it is integrated.

Bus interface: PCIe 3.0 x16 for the GT 640 versus Ring Bus for the UHD 610. Display outputs: 1x DVI, 1x HDMI 1.4a, 1x DisplayPort 1.2 for NVIDIA; motherboard dependent for Intel. Release dates differ by nearly eight years: 2012-06-04 for the GT 640, 2020-04-29 for the UHD 610. The GT 640 has a launch MSRP of 99 USD; the UHD 610 has no MSRP.

The Verdict

The data supports a clear split based on workload type. For raw compute performance in OpenCL and Vulkan, the NVIDIA GeForce GT 640 is the unambiguous choice. It wins both head-to-head tests by 38.8% and 35.5%, respectively, and its average score of 3,210 sits 16.6% above the UHD 610’s 2,752. Its 4x shading unit advantage, 8x ROP advantage, and dedicated 2 GB frame buffer with 28.51 GB/s bandwidth make it suitable for any application that needs consistent memory bandwidth and fill rate.

However, the Intel UHD Graphics 610 is not without merit. Its 15 W TDP is a fraction of the GT 640’s 65 W, and as an IGP it requires no separate card, no power connector, and no PSU headroom. It supports a higher DirectX feature level (12_1 versus 11_0) and a newer Vulkan version (1.3 versus 1.2.175). For systems where power efficiency and modern API support matter more than raw throughput, the UHD 610 is the sensible pick.

The GT 640’s nearest rivals show it is competitive within its own era: it beats the Arc Pro B60 by 0.9% and the Quadro P1000 by 1.5%, while trailing the GeForce 920M by 2.3% and GT 730M by 3.2%. The UHD 610 similarly trades blows with its peers, sitting 1.1% below the GeForce 820M and 2.3% above the RTX 5070 SUPER. Neither part is a performance leader, but the GT 640 is the stronger of the two in every measured benchmark.

Users who need the best possible OpenCL or Vulkan scores should pick the GT 640. Users who require a low-power integrated solution with modern DirectX 12.1 and Vulkan 1.3 support should pick the UHD 610. The GT 640’s performance lead is large, but the UHD 610’s architectural advantages in API support and power draw are equally measurable. The choice hinges on whether raw speed or efficiency and feature currency take priority.